DU145 (HTB-81), LNCaP (CRL-1740) and 293T (CRL-3216) cells were obtained in authenticated form (American Type Culture Collection, Manassas, VA) and cultured in DMEM plus 10% fetal bovine serum in a humidified atmosphere containing 5% CO₂ (21, 22). R427 DU145 cells were generated through homology-directed repair using the Alt-R CRISPR-Cas9 system and HDR donor oligos following methods recommended by the manufacturer (Integrated DNA Technologies, Coralville, IA). An AccuMax A302(IV) prostate cancer tissue microarray (Isu Abxis, Seongnam, Korea) encompassing matching normal and cancerous specimens was stained following standard procedures (23, 24) utilizing JMJD2D antibodies (ab93694; Abcam, Waltham, MA) at a dilution of 1:100.

Human 293T embryonic kidney cells were seeded onto poly-L-lysine coated dishes (25) and a day later transiently transfected by the calcium phosphate coprecipitation method (26, 27). Lysis of transfected cells and immunoprecipitation were then conducted as described before (28, 29). Immunoprecipitates were loaded onto SDS polyacrylamide gels and proteins blotted after separation onto polyvinylidene difluoride membranes (30). Proteins were detected with indicated primary antibodies and respective horseradish peroxidase-coupled secondary antibodies (31) employing enhanced chemiluminescence and exposure to film (32). The following primary antibodies were used: rabbit polyclonal p53-K372me₁ (ab16033; Abcam, Waltham, MA) and mouse monoclonal Flag M2 (F1804; Sigma-Aldrich, St. Louis, MO).

Hexahistidine-tagged SET8 and fusion proteins between glutathione S-transferase (GST) and JMJD2D or SET7/9 were expressed in Escherichia coli (33) and then purified utilizing Ni²⁺-nitrilotriacetic acid agarose or glutathione agarose, respectively (34). Proteins were incubated in 50 mM Tris-HCl (pH 8.5), 5 mM MgCl₂, 4 mM DTT at 30°C for 2 h in the presence of 1 µM ³H-labeled S-adenosyl-L-methionine. Thereafter, proteins were separated on SDS polyacrylamide gels (35), blotted onto polyvinylidene difluoride membrane (36) and visualized with Ponceau S staining (37). After drying, membranes were sprayed four times every 10 min with EN³HANCE (Perkin Elmer, Waltham, MA), dried again and exposed to film at -80°C (38).

Flag-JMJD2D was transiently coexpressed with Flag-SET7/9 in 293T cells and purified by anti-Flag immunoprecipitation (39). Further procedures were performed at the Mayo Clinic Proteomics Core Facility (Rochester, MN). Briefly, the immunoprecipitates were resolved on an SDS polyacrylamide gel, which was Coomassie stained, and the band corresponding to Flag-JMJD2D excised (40). After trypsin digestion, peptides were analyzed by MS/MS. Calculation of b, b+2H, y and y+2H ion masses was done through the following website: http://db.systemsbiology.net:8080/proteomicsToolkit/FragIonServlet.html; for methylation on K427 or K428, a mass of 14 was added to amino acid #8 or #9, respectively.

Retroviral expression vectors, which were based on the pQCXIH or pSIREN-RetroQ plasmids (Clontech, Mountain View, CA), were transfected together with packaging plasmids into 293T cells (41) and retrovirus collected 24 h and 48 h after transfection (42). Sequences of the utilized JMJD2D shRNAs have been published before (11); for CBLC, shRNA1 and shRNA2 targeted the sequences 5’-GCTGGCCATCATCTTCAGC-3’ and 5’-GTACTGTGGACACATGTAC-3’, respectively. Human DU145 prostate cancer cells were infected with respective retrovirus thrice every 12 h, grown for another 24 h, split and selected with 200 µg/ml hygromycin B or 1-2 µg/ml puromycin for 2-3 days (43). 2500 cells were then seeded into a well of a 96-well plate and cell growth measured with the PrestoBlue cell viability kit (Thermo Fisher, Waltham, MA) as the difference between the absorbance at 570 nm and 595 nm (44). Likewise, 2500 cells were seeded into a well of a 6-well plate and clonogenic activity determined after 10 days by counting crystal violet-stained colonies (45). For invasion assays, cells were treated with 10 µg/ml mitomycin C for 2 h and then their migration through 8 µm filters coated with Matrigel (BioCoat Growth Factor Reduced Matrigel Invasion Chambers #354483; Corning, Durham, NC) measured as described (46). Protein levels were assessed by Western blotting of the antibiotics-selected cells (47) utilizing rabbit polyclonal antibodies for JMJD2D (ab93694; Abcam, Waltham, MA) or mouse monoclonal antibodies for PLAGL1 (NBP2-37343; Novus Biologicals, Centennial, CO), while rabbit polyclonal actin antibodies (A2066; Sigma-Aldrich, St. Louis, MO) were used to control for comparable protein loading.

These assays were performed with transiently transfected 293T cells essentially as described (48, 49). Mouse monoclonal p53 (DO-1; sc-126; Santa Cruz Biotechnology, CA) or Flag M2 (F1804; Sigma-Aldrich, St. Louis, MO) antibodies were employed for immunoprecipitation, while mouse monoclonal Myc 9E10 (M4439; Sigma-Aldrich, St. Louis, MO) or Flag M2 antibodies were used for Western blotting (50).

Cells were crosslinked with 1% formaldehyde for 15 min, after which cells were lysed and sonicated and further processed as described before (51). For immunoprecipitations, rabbit IgG (sc-2027; Santa Cruz Biotechnology, CA) and rabbit polyclonal JMJD2D (ab93694; Abcam, Waltham, MA) antibodies were utilized. Finally, bound promoter fragments were amplified by nested PCR (52) and the resulting DNA fragments visualized through ethidium bromide staining after agarose electrophoresis (53). The PCR primers used were: CBLCfor-401 (5’-GTAGAGACACGGTTTCACCATGTTGG-3’), CBLCfor-316 (5’-CTGGGATTACAGTTGTGAGTCATCGC-3’), CBLCrev-61 (5’-TGCAGGTACCAGTGTCTCCAAAGGGG-3’), CBLCrev-10 (5’-CTCGCCCAGAGTGAAAGGAGAGG-3’), PLAGL1for-434 (5’-CGTTTCTCATGTGTGATTGGGCTCTG-3’), PLAGL1for-398 (5’-CTGGCGGAGACTTCGGCTAGCAGG-3’), PLAGL1rev-49 (5’-GACGGGCTGAATGACAAATGGCAG-3’) and PLAGL1rev-15 (5’- CAGCCGTGTCTAAATCAAGGCTCG-3’).

RNA was isolated employing Trizol as described (54). After reverse transcription with random primers, RNA was amplified by PCR (55) and relative gene expression was determined with the ΔΔCt method and normalization to levels of GAPDH (52). Sequences of oligonucleotides utilized for PCR are listed in Supplementary Methods. RNA sequencing was performed by Novogene (Sacramento, CA) and data analyzed as described before (56).

5-week-old immunocompromised nude male mice (Fox1^nu/Fox1^nu ; #007850; Jackson Laboratory, Bar Harbor, ME) were acclimated for two weeks and then subcutaneously injected through a 27 gauge needle into the right flank (57). For this, 2x10⁶ cells resuspended in 100 µl PBS plus 100 µl growth-factor reduced Matrigel (#354230; Corning, Durham, NC) were employed. Tumor volume (calculated as 0.5 x width x width x length) was measured weekly and tumors were dissected after euthanasia to determine their weight. These mouse experiments were approved by the University of Oklahoma Health Sciences Center Institutional Animal Care and Use Committee.

Previously, we found that JMJD2D binds to the tumor suppressor p53 (11). It is known that p53 is regulated by methylation, including through SET7/9-mediated methylation of its K372 residue (58). Hence, we wondered if the JMJD2D demethylase would oppose SET7/9-mediated methylation of p53. While we found no evidence for that, we serendipitously uncovered – upon using an antibody recognizing p53 monomethylated on K372 – that JMJD2D itself became methylated when SET7/9 was overexpressed in 293T cells ( Figure 1A ). Furthermore, deleting 169 C-terminal amino acids (see ΔC (2–354) truncation in Figure 1A ) abrogated this methylation in JMJD2D, suggesting that a lysine residue(s) close to the C-terminus of JMJD2D became methylated by SET7/9.

To demonstrate that SET7/9 directly methylates JMJD2D, we purified respective recombinant proteins and performed in vitro methylation assays with radioactive S-adenosyl-L-methionine as a methyl donor. SET7/9 methylated p53 as expected ( Figure 1B , left panels) and also JMJD2D ( Figure 1B , right panels). Moreover, another methyltransferase, SET8, which has been reported to methylate p53 (59), did not utilize JMJD2D as a substrate, whereas a weak activity towards p53 was detectable. These data indicate that SET7/9 can directly methylate JMJD2D.

To identify the site(s) of SET7/9-mediated methylation in JMJD2D, we coexpressed both proteins in 293T cells and then performed mass spectrometry on immunopurified JMJD2D. This revealed only one tryptic peptide consisting of JMJD2D amino acids 420-450 that became monomethylated. This peptide encompassed two lysine residues (K427 and K428). However, the observed fragmentation pattern was only consistent with methylation on K427 ( Figure 2 ; see ions b8 and y23 + 2H).

To validate this inference, we also generated mutated GST-JMJD2D fusion proteins and examined their ability to become methylated by SET7/9 in vitro. In contrast to wild-type JMJD2D, the K427R mutant was no longer methylated ( Figures 3A, B ). As a control, a K428R mutant was still utilized as a substrate by SET7/9, while a K427R/K428R double mutant was not. Similarly, we assessed in vivo methylation with JMJD2D point mutants. Again, we observed that mutation of K427, but not K428, to arginine compromised SET7/9-dependent methylation ( Figure 3C ). Please note that amino acids surrounding JMJD2D-K427 and p53-K372 are similar ( Figure 3D ), potentially explaining why the antibody raised against monomethylated K372 of p53 cross-reacted with JMJD2D when monomethylated on K427. Altogether, our data demonstrate that K427 in JMJD2D can be methylated by SET7/9.

To examine a potential role of JMJD2D in prostate cancer, we first stained a human tissue microarray composed of matching normal and cancerous prostate tissue with JMJD2D antibodies in order to answer if JMJD2D is expressed in the human prostate. Indeed, JMJD2D expression was observed in both normal prostates and adenocarcinomas; notably, its expression was significantly enhanced in tumors ( Figures 4A, B ). Further, downregulation of JMJD2D with two different shRNAs resulted into less growth and clonogenic activity of DU145 human prostate cancer cells ( Figures 4C–E ), and similarly JMJD2D downregulation diminished the growth of human LNCaP prostate cancer cells ( Supplementary Figure 1 ). These results suggest that JMJD2D is a promoter of prostate tumorigenesis.

To determine how methylation of K427 would affect JMJD2D’s function, we examined if the R427 mutation would tamper the interaction of JMJD2D with two of its prostate cancer-relevant interaction partners, the tumor suppressor p53 and the ETS protein ETV1 (11, 19). However, wild-type and R427 JMJD2D did not noticeably differ in their abilities to form complexes with p53 or ETV1 ( Supplementary Figure 2 ). We also performed CRISPR/Cas9-mediated mutagenesis in DU145 prostate cancer cells in order to replace K427 by an arginine residue and were able to obtain respective homozygous R427 DU145 cells ( Supplementary Figures 3A, B ). Notably, this did not lead to a change in H3K9me₃ levels ( Supplementary Figure 3C ). Then, we compared wild-type to R427 DU145 prostate cancer cells and found that R427 cells displayed significantly reduced growth, clonogenic activity and invasion potential in vitro ( Figures 5A–C ). Further, we injected these cells subcutaneously into nude mice and observed that the R427 cells were also compromised in their ability to form tumors in vivo ( Figures 5D, E ). Collectively, these data indicate that K427 methylation stimulates JMJD2D’s oncogenic properties in DU145 cells.

To understand how mutation of K427 affects DU145 cell physiology, we performed RNA sequencing. Compared to wild-type DU145 cells, the R427 cells displayed roughly each 400 significantly up- and downregulated genes ( Figure 6A ). From those, we selected 8 genes (CBLC, METTL27, COL4A5, GLIS3, NPR1, OSR2, PLAGL1, RSPO3) based on their known or their unanticipated function in prostate cancer (see below and Discussion) and corroborated their altered expression by quantitative RT-PCR ( Figure 6B ). Bioinformatic analyses ( Supplementary Figure 4A ) indicated that genes upregulated in R427 cells (COL4A5, GLIS3, NPR1, OSR2, PLAGL1, RSPO3) were all significantly downregulated in human prostate adenocarcinomas, while genes downregulated in R427 cells were either significantly higher expressed (CBLC) or trending towards enhanced expression (METTL27) in prostate tumors. This would be consistent with the notion that methylation of JMJD2D could promote prostate cancer development through upregulation of CBLC and METTL27 as well as through downregulation of COL4A5, GLIS3, NPR1, OSR2, PLAGL1 and RSPO3.

Our interest was particularly piqued by CBLC (cbl proto-oncogene C) since it is robustly expressed in normal prostate tissue (60, 61), yet its relationship with prostate cancer has remained unknown. Moreover, chromatin immunoprecipitation experiments indicated that JMJD2D interacted with the CBLC gene promoter ( Supplementary Figure 4B ), strongly suggesting that JMJD2D can directly regulate CBLC transcription. To study the role of CBLC in DU145 cells, we downregulated it with two different shRNAs ( Figure 7A ). This resulted into significantly reduced cell growth, clonogenic activity, invasion and tumor formation ( Figures 7B–F ), which is similar to the phenotype of the R427 mutation (see Figure 5 ). Our data have thereby identified CBLC as a novel promoter of prostate cancer and to be potentially one of the seminal genes regulated by methylated JMJD2D.